Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Advertisement
Humanities and Social Sciences Communications volume 13, Article number: 1716 (2026)
1230
Metrics details
This study examines how chatbot interactivity and limitation transparency shape users’ receptiveness to chatbot recommendations in environmental communication. Drawing on the theory of interactive media effects, we conceptualize these two chatbot features as distinct design cues that influence message acceptance through different psychological mechanisms. Across three experimental studies with Taiwanese participants, the findings show that chatbot interactivity primarily increases receptiveness by strengthening perceived integrity, whereas limitation transparency enhances receptiveness by reducing skepticism. A factorial experiment further demonstrates that the combination of high interactivity and high limitation transparency produces the strongest effects by simultaneously increasing perceived integrity and reducing skepticism. These findings extend interactive media research by showing that, in environmental communication, responsiveness and transparent disclosure operate as complementary trust-relevant cues. The study also offers practical guidance for designing chatbots for sustainability communication by highlighting the importance of combining engaging interaction with explicit acknowledgment of system limitations.
The rapid advancement of artificial intelligence (AI) has opened new avenues for promoting sustainability. Chatbots are increasingly used among AI technologies to deliver sustainability messages (Johnson 2023). Organizations such as the World Wildlife Fund and Ecosia have implemented AI chatbots to promote conservation and reforestation (Johnson 2023; World Wildlife Fund 2024). However, sustainability communication is a challenging context for chatbot deployment, as users cannot easily verify environmental claims, are sensitized to greenwashing and persuasive manipulation, and must rely heavily on source and process cues to evaluate credibility (Endsley 2023). This raises critical questions about how chatbot design features shape the perceptions that underpin receptiveness to AI-generated environmental guidance.
The theory of interactive media effects suggests that users are more likely to trust and accept interactive media due to their dynamic (i.e., tailored to user inputs or demands) two-way communication (Wang and Sundar 2018). For chatbots, two key dimensions—interactivity (how responsive and engaging users perceive the chatbot to be) and limitation transparency (disclosure of the chatbot’s capabilities and boundaries)—are crucial in shaping the user experience (Cheng et al. 2024; Hadi et al. 2024). These dimensions influence trust and engagement, which are vital in environmental communication (Lewis et al. 2022). In environmental communication, trust formation is cue-driven because audiences often cannot easily validate sustainability claims and therefore rely more on source and process cues to judge credibility and intent. Accordingly, when a chatbot both engages users through contingent dialogue and signals epistemic humility about what it can and cannot reliably infer, users should be less likely to interpret sustainability guidance as overconfident, opportunistic, or greenwashing-adjacent (Mozafari et al. 2021; Endsley 2023; Lee and Raschke 2023; Binesh et al. 2024; Chen et al. 2022). Receptiveness is a key driver of sustainable practice adoption (Aljukhadar et al. 2010). While prior research has highlighted the role of interactivity and transparency in general technology acceptance, it remains unclear how these features influence receptiveness to chatbot-generated environmental messages. This is especially important given users’ concerns about whether AI can provide personalized, accurate sustainability guidance (Endsley 2023). Accordingly, this research investigates whether chatbot interactivity and limitation transparency can enhance environmental message acceptance.
Given that in persuasive domains such as environmental communication, high interactivity alone may also raise concerns about strategic persuasion or impression management, thereby limiting its positive effects (Lewis et al. 2022; Mozafari et al. 2021). This risk is amplified in environmental contexts, where audiences are sensitized to persuasive manipulation and reputation management amid widespread concerns about greenwashing. Limitation transparency may function as a critical moderating cue because it directly shapes users’ inferences about the chatbot’s honesty and epistemic limits (Endsley 2023). By openly acknowledging its constraints, the chatbot reduces suspicions of exaggeration or manipulation and legitimizes the persuasive influence of interactivity (Chen et al. 2022; Endsley 2023). When both interactivity and transparency are high, users are more likely to interpret interactive engagement as genuine rather than strategic, thereby strengthening the effects of interactivity on perceived integrity, skepticism, and receptiveness.
Another crucial aspect of effective sustainability communication refers to users’ psychological responses to the message source, particularly perceived integrity and skepticism. Integrity reflects belief about the chatbot’s honesty, accuracy, and ethicality, whereas skepticism reflects doubt about its credibility and intentions (Xie and Peng 2009; Yin et al. 2022). These evaluations are highly salient in an era of misinformation and greenwashing (Lee and Raschke 2023; Santos et al. 2024). When chatbots exhibit high interactivity and openly communicate their limitations, they signal transparency and responsiveness—cues associated with users’ perceptions of the chatbot’s integrity and reduce doubts about manipulation or deception (Cheng et al. 2024; Hadi et al. 2024). These evaluations in turn influence whether users trust and follow chatbot recommendations: high integrity builds credibility, while low skepticism reduces resistance, both increasing message receptiveness (Hadi et al. 2024; Mozafari et al. 2021; Pizzi et al. 2021). Notably, these evaluations are not necessarily mutually inclusive, as prior research indicates that whether a communicator is manipulative (i.e., integrity) or not does not necessarily shape the believability of the information they relay (i.e., skepticism) (Pizzi et al. 2021; Yin et al. 2022). This research draws on this knowledge to examine the mediating roles of perceived integrity and chatbot skepticism in the effects of chatbot interactivity and transparency on user receptiveness to chatbot recommendations.
Through a series of three experiments, this research addresses two central questions: (1) whether chatbot message interactivity and limitation transparency enhance user receptiveness to chatbot-generated recommendations in sustainability communication, and (2) whether these effects are mediated by perceived integrity and skepticism. To examine these questions, three experimental studies were conducted with Taiwanese participants. The findings offer two key contributions to the literature on artificial intelligence-mediated communication. First, the study extends the theory of interactive media effects by theorizing and empirically demonstrating that message interactivity and limitation transparency operate as trust-relevant design cues in environmental communication, a domain characterized by low claim verifiability and heightened greenwashing risk. In such contexts, users are more likely to rely on integrity- and motive-based inferences when evaluating AI-generated messages (Wang and Sundar 2018; Lewis et al. 2022; Lee and Raschke 2023; Santos et al. 2024). Second, the results show that these design cues influence receptiveness specifically through psychological evaluations that are central to sustainability persuasion, namely perceived integrity and skepticism. By identifying these mechanisms, the study clarifies how AI systems can gain acceptance in contested, credibility-sensitive environmental messaging environments. Whereas prior research on AI interactivity has primarily examined outcomes such as engagement, credibility, and persuasion in health and commercial contexts (Hu and Wise 2021; Lou et al. 2022), the present findings situate these psychological processes within environmental communication, where skepticism toward message authenticity and concerns about deceptive intent are more pronounced.
The theory of interactive media effects proposes that interactive communication environments foster greater engagement, trust, and message processing by enabling dynamic, reciprocal exchanges between communicators and users (Wang and Sundar 2018). Beyond enhancing involvement and perceived control, such exchanges serve as diagnostic design cues that enable users to infer a communicator’s credibility, intent, and reliability (Lee et al. 2021). This perspective is relevant for AI-based sustainability communication, where audiences often cannot directly verify environmental claims and must rely on interactional and procedural cues to judge trustworthiness and persuasive intent (Lee and Raschke 2023). In this context, interactive media effects complement established frameworks such as source credibility theory and transparency theory by explaining how communication process features, rather than message content alone, shape integrity and motive inferences (Hu and Wise 2021; Lou et al. 2022). At the same time, they extend persuasion knowledge perspectives by specifying how contingent responsiveness and system disclosure operate as cues that either legitimize or undermine persuasive influence in algorithmic communication environments (Følstad et al. 2018; Pizzi et al. 2021). Importantly, interactive media effects emphasize that persuasion outcomes depend not only on interface affordances but also on the contingent nature of dialogue, making this framework especially well-suited for examining conversational AI systems in credibility-sensitive domains such as environmental communication.
Prior research distinguishes between modality interactivity and message interactivity. Whereas modality interactivity concerns interface-level affordances such as navigational control or multimedia features, message interactivity reflects the degree to which system responses are meaningfully contingent on users’ prior inputs (Wang and Sundar 2018; Pizzi et al. 2021). This contingency-based view emphasizes reciprocity, responsiveness, and continuity across conversational turns as the primary mechanism through which interactivity shapes psychological outcomes.
In the context of AI chatbots, interactivity is commonly conceptualized as perceived responsiveness and adaptive dialogue (Hadi et al. 2024). Building on this perspective, the present research defines interactivity as message interactivity, that is, the extent to which a chatbot’s replies are responsive and meaningfully adapted to users’ preceding inputs, enabling a coherent two-way exchange (Lou et al. 2022). Because chatbot communication is inherently conversational rather than interface-driven, message interactivity represents the most theoretically relevant dimension for predicting trust-related inferences in AI-mediated interactions. In environmental communication, where users often confront uncertainty, contested claims, and complex behavioral recommendations, such contingent dialogue serves as a diagnostic for assessing whether sustainability advice is personalized, sincere, and offered in good faith.
In contrast to interactivity, limitation transparency represents a distinct design attribute reflecting the chatbot’s proactive disclosure of its epistemic and functional boundaries. Rather than constituting a component of interactivity, transparency operates as a signaling mechanism that communicates honesty, accountability, and epistemic humility (Endsley 2023; Mozafari et al. 2021). When users perceive that an AI system openly communicates its limitations, they are more likely to view its recommendations as credible and less strategically motivated (Chen et al. 2022). This function is particularly salient in environmental communication, where misinformation and greenwashing frequently undermine public trust and increase scrutiny of persuasive claims (Lewis et al. 2022; Santos et al. 2024).
Accordingly, the present study treats chatbot interactivity and limitation transparency as theoretically distinct yet complementary design cues that jointly shape users’ receptiveness to chatbot recommendations in environmental communication.
Environmental communication typically concerns contested issues that require audiences to invest substantial cognitive and moral effort to evaluate claims and recommendations (Xie and Peng 2009). In such contexts, perceived integrity of the message source plays a central role in shaping persuasion outcomes. Perceived integrity reflects judgments about a communicator’s honesty, reliability, and ethical standing (Yin et al. 2022). When message sources are perceived as acting in good faith and aligning with ethical principles, audiences are more inclined to trust and accept pro-environmental appeals (Jiang et al. 2023). Conversely, sustainability messages often fail when audiences suspect strategic motives, public relations agendas, or exaggeration.
Closely related is skepticism, which reflects doubt regarding the credibility, accuracy, or intentions underlying persuasive communication (Ahadzadeh et al. 2021). In environmental domains, skepticism is frequently amplified by greenwashing, unverifiable impact claims, and scientific uncertainty (Lee and Raschke 2023; Santos et al. 2024). Under such conditions, users may defensively discount recommendations even when they are well-intentioned. Because environmental advice often involves long-term and probabilistic outcomes, users rely heavily on source cues when forming receptiveness judgments.
In the present research context, skepticism does not primarily concern suspicions of corporate self-interest or strategic greenwashing, as the chatbot is presented as a general language model rather than an organizational agent. Instead, skepticism reflects doubts about the accuracy, completeness, and potential overstatement of AI-generated sustainability advice (Lee and Raschke 2023). Given growing public awareness of AI hallucination, probabilistic inference, and overconfident system outputs, users may question whether chatbot recommendations are reliable, overly generic, or presented with unwarranted certainty, even in the absence of explicit organizational motives (Endsley 2023; Schmidt et al. 2020). Thus, skepticism in this study captures epistemic and informational doubts about the quality and trustworthiness of AI-generated environmental guidance rather than suspicions of institutional persuasion.
For AI chatbots, perceived integrity and skepticism constitute core determinants of trust, a construct widely recognized as mediating the relationship between source attributes and persuasion outcomes (Pizzi et al. 2021). When a chatbot is perceived as honest and free from manipulative intent, users are more likely to accept its recommendations (Følstad et al. 2018). Conversely, suspicions of bias or strategic persuasion can undermine receptiveness even when message content is reasonable (Chaves and Gerosa 2021). Transparent disclosure of a chatbot’s limitations may therefore alleviate concerns about misinformation and opportunistic influence, strengthening integrity judgments and reducing skepticism (Binesh et al. 2024).
Importantly, perceived integrity and skepticism are conceptually distinct and not redundant. Low skepticism does not necessarily imply high perceived integrity: users may withhold doubt while remaining neutral about the communicator’s ethical character (Pizzi et al. 2021). Likewise, users may perceive a chatbot as sincere yet remain skeptical about the adequacy or applicability of its recommendations under uncertainty (Yin et al. 2022). Accordingly, the present research treats perceived integrity and skepticism as parallel psychological mechanisms (one affirming and evaluative, the other defensive and doubt-based) that jointly shape receptiveness to chatbot recommendations in environmental communication. Drawing on the theory of interactive media effects, the present research proposes that perceived integrity and skepticism constitute central psychological mechanisms through which chatbot interactivity and limitation transparency shape receptiveness to chatbot recommendations in environmental communication.
Chatbot interactivity fosters user engagement through personalized and reciprocal communication, signaling that the chatbot is invested in providing tailored solutions rather than generic advice (Chen et al. 2022; Lou et al. 2022). Users experiencing such personalized communication are likely to perceive recommendations as more relevant and valuable (Hadi et al. 2024). Additionally, interactivity reduces perceptions of manipulative intent by demonstrating sincerity and adaptive communication rather than promotional messaging (Pizzi et al. 2021), thereby increasing users’ receptiveness to chatbot recommendations. We propose the following hypothesis:
H1: Chatbot interactivity positively affects receptiveness to chatbot recommendations.
Similarly, limitation transparency demonstrates honesty and accountability by explicitly communicating knowledge boundaries (Valencia et al. 2023). This openness diminishes uncertainty about the chatbot’s motivations and reduces suspicion of hidden biases, thereby increasing perceived authenticity and reliability (Binesh et al. 2024; Endsley 2023). In environmental communication contexts where credibility is paramount, transparency about limitations encourages users to view recommendations as trustworthy, particularly given that sustainability advice often involves uncertain outcomes and contested claims (Lewis et al. 2022; Mozafari et al. 2021).
H2: Chatbot limitation transparency positively affects receptiveness to chatbot recommendations.
Beyond openness, both interactivity and transparency shape users’ perceptions of chatbot integrity. Interactive features signal not only responsiveness but also benevolent intent and procedural fairness (Wang and Sundar 2018). When chatbots dynamically respond to user input, they convey a willingness to prioritize user needs rather than pursuing predetermined agendas (Hadi et al. 2024; Sun et al. 2024). This contingent responsiveness allows users to observe how the chatbot adapts based on their input, reducing suspicions of scripted persuasion and signaling fairness (Lin 2024). By demonstrating procedural responsiveness, interactive chatbots foster perceptions of transparent, ethical operation without manipulative intent (Lou et al. 2022).
H3: Chatbot interactivity positively affects perceived integrity.
Limitation transparency similarly enhances integrity by signaling openness and honesty (Valencia et al. 2023). When chatbots acknowledge knowledge boundaries, they reassure users that they are not attempting to mislead or fabricate information (Endsley 2023). This transparency reduces perceptions of opportunistic intent and reinforces beliefs that the chatbot operates ethically and without bias (Binesh et al. 2024). Research confirms that communicators who proactively disclose limitations are perceived as more trustworthy and authentic (Lee and Johnson 2022).
H4: Chatbot limitation transparency positively affects perceived integrity.
These design features also combat skepticism, which arises when communication appears closed, manipulative, or opaque (Ahadzadeh et al. 2021; Su et al. 2021). Interactivity reduces skepticism by creating personalized engagement and responsiveness. When chatbots adapt responses based on user inputs, they signal attentiveness and reduce perceptions of agenda-driven messaging (Hadi et al. 2024; Pizzi et al. 2021). The reciprocity of interactive communication fosters dialogue and transparency, making users more likely to perceive the chatbot as reliable (Lin 2024; Lou et al. 2022).
H5: Chatbot interactivity negatively affects skepticism.
Limitation transparency similarly mitigates skepticism. When chatbots disclose constraints, such as indicating insufficient information or probabilistic recommendations, users perceive them as more credible and less likely to provide misleading claims (Chen et al. 2022; Mozafari et al. 2021). This openness aligns with honesty expectations and reduces suspicions of hidden biases (Binesh et al. 2024). By proactively revealing constraints, chatbots demonstrate accountability and diminish credibility doubts (Følstad et al. 2018).
H6: Chatbot limitation transparency negatively affects skepticism.
The consequences of perceived integrity and skepticism for recommendation receptiveness are critical. Perceived integrity fosters trust by reinforcing beliefs that chatbots provide unbiased, ethically sound advice (Pizzi et al. 2021), making users more likely to accept information and adopt suggested behaviors (Jiang et al. 2023). In AI-driven environmental communication, integrity perceptions increase willingness to trust sustainability advice and facilitate engagement with pro-environmental content (Hadi et al. 2024; Mozafari et al. 2021).
H7: Perceived integrity positively affects receptiveness to chatbot recommendations in environmental communication.
Conversely, skepticism leads users to critically scrutinize and resist AI-generated recommendations (Lee et al. 2021). Users harboring doubts about AI systems are more likely to dismiss recommendations, viewing them as potentially biased, misleading, or driven by hidden agendas (Jiang et al. 2023; Mozafari et al. 2021).
H8: Skepticism negatively affects receptiveness to chatbot recommendations in environmental communication.
Beyond these direct effects, perceived integrity and skepticism serve as mediating mechanisms. Interactivity enhances user engagement through dynamic, responsive communication (Hadi et al. 2024), and users attribute higher attentiveness to interactive systems (Lou et al. 2022; Pizzi et al. 2021). By facilitating reciprocal exchanges, interactivity reduces perceptions of generic content delivery, enhancing credibility (Sun et al. 2024). Users view interactive chatbots as operating with fairness and integrity because they appear invested in understanding needs rather than disseminating predetermined messages (Lin 2024). This perceived integrity enhances trust in recommendations, making users more receptive to sustainability guidance (Mozafari et al. 2021). Without high integrity, users may question whether chatbots provide genuinely unbiased information or merely use engagement tactics to push agendas (Lou et al. 2022).
H9a: Perceived integrity mediates the effects of chatbot interactivity on receptiveness to chatbot recommendations.
Similarly, limitation transparency fosters confidence in chatbot honesty (Binesh et al. 2024). Users perceive systems as acting in good faith when they openly communicate limitations rather than projecting infallibility (Valencia et al. 2023). Disclosing knowledge boundaries reduces suspicions of manipulative intent (Chen et al. 2022), increasing willingness to trust AI-generated recommendations (Mozafari et al. 2021). Without high integrity, even transparent chatbots may face resistance due to suspected hidden biases (Mozafari et al. 2021).
H9b: Perceived integrity mediates the effects of chatbot limitation transparency on receptiveness to chatbot recommendations.
Skepticism also mediates these relationships. Interactivity enables dynamic, two-way dialogue rather than one-directional information delivery (Pizzi et al. 2021). In environmental communication, skepticism often arises when messages feel generic or pre-scripted (Sun et al. 2024). Interactive chatbots counteract this by allowing users to ask questions, request clarifications, and receive tailored responses, making communication feel responsive to concerns (Lin 2024). This reduces skepticism and fosters openness to sustainability recommendations (Chaves and Gerosa 2021). However, if interactivity appears superficial or manipulative, skepticism may persist (Mozafari et al. 2021).
H10a: Skepticism mediates the effects of chatbot interactivity on receptiveness to chatbot recommendations.
Finally, limitation transparency counteracts skepticism by signaling honesty and reducing perceived bias (Binesh et al. 2024). In environmental communication, where misinformation is concerning, users trust chatbots that admit uncertainty rather than presenting complete knowledge illusions (Chen et al. 2022). By explicitly stating when data is incomplete or recommendations are probabilistic, chatbots reduce deception perceptions and lower skepticism (Hadi et al. 2024), making users more receptive to environmental recommendations (Følstad et al. 2018).
H10b: Skepticism mediates the effects of chatbot limitation transparency on receptiveness to chatbot recommendations.
Interactivity and limitation transparency represent conceptually distinct design attributes that need not covary and may coexist in practice. A chatbot may exhibit high interactivity by providing highly contingent and adaptive responses while simultaneously being transparent about the limits of its knowledge or personalization capacity (Pizzi et al. 2021; Hadi et al. 2024). In contemporary AI systems, fluent and responsive chatbots often operate under epistemic uncertainty and may generate probabilistic or incomplete recommendations, a phenomenon associated with overconfidence and hallucination in AI outputs (Endsley 2023; Schmidt et al. 2020). In such cases, high interactivity does not imply high epistemic reliability. Limitation transparency, therefore, does not negate interactivity; rather, it qualifies the scope and certainty of interactive responses by clarifying that recommendations are general or subject to error (Chen et al. 2022; Valencia et al. 2023). This distinction allows chatbots to remain conversationally responsive while honestly communicating their epistemic boundaries, thereby shaping users’ inferences about credibility and intent (Mozafari et al. 2021; Wanner et al. 2022). Accordingly, interactivity and limitation transparency can vary independently and jointly shape how users evaluate chatbot trustworthiness.
When chatbots simultaneously exhibit high interactivity and high limitation transparency, users are more likely to perceive them as honest, ethical, and unbiased (Binesh et al. 2024; Endsley 2023; Mozafari et al. 2021). In contrast, low interactivity fosters impressions of generic and scripted communication (Sun et al. 2024), while low limitation transparency undermines perceptions of honesty and accountability (Chen et al. 2022; Valencia et al. 2023). The joint presence of high interactivity and transparency, therefore, uniquely enhances perceived integrity by signaling personalized attention, sincerity, and ethical conduct (Xie and Peng 2009), while reducing suspicions of hidden motives (Goh and Balaji 2016). This enhanced integrity is expected to increase users’ receptiveness to chatbot recommendations.
Similarly, the combination of high interactivity and high limitation transparency is expected to reduce skepticism. Interactivity mitigates skepticism by signaling responsiveness and genuine consideration of user needs (Chaves and Gerosa 2021; Hadi et al. 2024), whereas limitation transparency reduces uncertainty and alleviates concerns about hidden biases or misinformation (Binesh et al. 2024; Chen et al. 2022). Conversely, low interactivity and low transparency intensify perceptions of manipulation and exaggeration (Mozafari et al. 2021). Thus, the synergistic presence of high interactivity and high limitation transparency should jointly reduce skepticism and enhance openness to chatbot-delivered environmental recommendations. The following hypotheses are therefore proposed:
H11a: The interaction between interactivity and limitation transparency increases receptiveness through perceived integrity.
H11b: The interaction between interactivity and limitation transparency increases receptiveness through reduced skepticism.
Figure 1 illustrates the proposed research model depicting the relationships among chatbot interactivity, limitation transparency, perceived integrity, skepticism, and receptiveness to chatbot recommendations in environmental communication.
The figure presents the hypothesized relationships among chatbot interactivity, limitation transparency, perceived integrity, skepticism, and receptiveness to chatbot recommendations in environmental communication.
The research program comprises three experimental studies designed to systematically examine the effects of chatbot interactivity and limitation transparency on user receptiveness and the underlying psychological mechanisms. Study 1 examines chatbot interactivity, Study 2 isolates limitation transparency, and Study 3 integrates both attributes in a factorial design to test their joint effects.
Across all studies, vignette-based chatbot interaction scenarios were employed to ensure consistency and comparability. Each experimental condition used a single standardized scenario to maintain internal validity and to ensure that observed effects could be attributed to the manipulated design attributes rather than to content variation. The recommendations reflected commonly encountered environmental guidance to enhance ecological validity and simulate realistic chatbot-based sustainability interactions. The scenarios followed a consistent structure consisting of a user query, a chatbot recommendation, and the embedded manipulation. Only interactivity and/or limitation transparency varied across conditions, while the substantive content remained constant. Although the scenarios did not involve extended multi-turn dialogue, this format was sufficient to capture message interactivity through cues of responsiveness and contingency in the chatbot’s reply. Specifically, interactivity was conveyed through the extent to which the chatbot engaged with the user’s query and tailored its response to the user’s expressed request, which represents a controlled yet theoretically meaningful operationalization of message interactivity in vignette-based experimentation. Detailed experimental scenarios and measurement items are provided in the Appendix section (Appendices 1–9), with measurement scales reported in Appendix 9. Each appendix corresponds to a specific experimental condition or measurement set, ensuring full transparency and reproducibility.
Study 1 examined the direct effects of chatbot interactivity on perceived integrity, perceived skepticism, and receptiveness to chatbot recommendations. It also aimed to examine the indirect effects of perceived chatbot interactivity on receptiveness to chatbot recommendations. The study employed a between-subjects experimental design. Prior to the main study, a pretest (n = 64) was conducted to confirm that the vignette scenarios clearly conveyed the intended difference in chatbot interactivity and were understandable to participants. The pretest results supported the adequacy of the manipulation, showing that the high-interactivity scenario was perceived as significantly more interactive than the low-interactivity scenario (Mhigh = 4.62, SD = 1.01 vs. Mlow = 3.28, SD = 1.14, t(78) = 5.56, p < 0.001). In addition, participants rated the scenarios as sufficiently clear and realistic (Mclarity = 5.21, SD = 0.88; Mrealism = 4.97, SD = 0.93), providing further support for the use of the stimuli in the main experiment.
For the main study, a total of 600 participants were recruited from Taiwan (57.7% female, 44.2% aged 41–55, 38.7% married) via an online platform and were compensated with a convenience store gift card for their participation. They were randomly assigned to one of two experimental conditions: low chatbot interactivity or high chatbot interactivity.
In both conditions, participants were shown a vignette depicting a hypothetical interaction with a chatbot. They were asked to imagine that they asked a language model to provide information on how they could be more environmentally friendly. In the low interactivity condition, the chatbot provided a direct answer, indicating how they could be more environmentally friendly without providing room for any further clarification by the individual. In the high interactivity condition, before providing the answer, the chatbot asked the customer about any eco-friendly habits they were already practicing. After the user’s response, the chatbot provided suggestions on how the individual could be more environmentally friendly and then told the customer to reach out to it if they had any more questions. The scenarios are presented in Appendices 1 and 2. Thereafter, the participants responded to manipulation check items for chatbot interactivity (M = 3.99, SD = 1.36, α = 0.965; Lee et al. 2021), followed by questions measuring perceived integrity (M = 5.05, SD = 0.67, α = 0.758), skepticism (M = 3.75, SD = 1.08, α = 0.926), and receptiveness to chatbot recommendations (M = 4.45, SD = 1.00, α = 0.939), all measured on the 7-point Likert Scale (see Appendix 9 for detailed items and their sources). At the end of the survey, demographic information was collected.
To verify the effectiveness of the experimental manipulations, manipulation checks were conducted using independent samples t-tests. Results indicated that participants in the low interactivity condition rated the chatbot as significantly less interactive compared to those in the high interactivity condition (MLowINT = 2.65, MHighINT = 5.32, t = −23.50, p < 0.001). This indicates that the manipulations were successful.
To examine both direct and indirect (mediating) effects, the SPSS PROCESS Macro (version 4.2) was employed. A custom model was developed based on the procedures outlined by Hayes (2022). A dummy variable for interactivity (0 = low interactivity, 1 = high interactivity) was used as the independent variable. The results (see Table 1) confirmed that high interactivity was associated with higher receptiveness to chatbot recommendations (β = 0.236, p < 0.001, 95% CI = [0.186, 0.302]), higher perceived integrity (β = 0.390***, CI = [0.140, 0.640]), and lower skepticism (β = −0.491***, CI = [−0.132, −0.649]), supporting H1, H3, and H5. Moreover, the results confirmed the positive effects of perceived integrity on receptiveness to chatbot recommendations (β = 0.074*, CI = [0.032, 0.118]), supporting H7. The mediating role of perceived integrity in the relationship between chatbot interactivity and receptiveness to chatbot recommendations was also confirmed (β = 0.029, CI = [0.017, 0.160]), supporting H9a. On the other hand, the effect of skepticism on receptiveness to chatbot recommendations was not significant (β = −0.001, CI = [−0.069, 0.069]). Consequently, the mediating effect of skepticism was not significant (β = 0.001, CI = [−0.138, 0.145]). Thus, H8 and H10a were not supported.
Study 1 confirmed that interactive chatbot messages enhance receptiveness to recommendations. Specifically, the results demonstrated that, compared to low interactivity, high interactivity significantly increased participants’ receptiveness to chatbot recommendations in the context of environmental communication. This finding aligns with prior research suggesting that interactive exchanges between chatbots and users foster the perception that the chatbot is not delivering generic responses, but is instead attempting to understand the user’s unique situation and provide tailored advice (Hadi et al. 2024; Mozafari et al. 2021; Sun et al. 2024). Moreover, the results indicate that the effect of chatbot message interactivity on receptiveness is mediated by perceived integrity. Interactive chatbots appear to reduce users’ perceptions of opportunistic motives by signaling openness and a willingness to engage in meaningful dialogue. This perception of fairness arises from the chatbot’s ability to respond thoughtfully rather than delivering biased or scripted messages (Lin 2024; Lou et al. 2022). Contrary to expectations, skepticism did not significantly influence receptiveness to chatbot recommendations, nor did it mediate the relationship between message interactivity and receptiveness. Some scholars suggest that in the context of environmental communication, skepticism functions as a peripheral cue (Goh and Balaji 2016). When chatbot-delivered information requires more cognitive effort, such as in environmental topics, peripheral cues like skepticism may play a less prominent role in shaping user responses (Goh and Balaji 2016; Su et al. 2021).
While Study 1 demonstrates that chatbot interactivity enhances receptiveness through increased perceived integrity, it focuses exclusively on one design attribute, namely message interactivity, and does not consider how other design cues may independently shape user trust and resistance. In particular, Study 1 does not address how users respond when chatbots explicitly communicate the limits of their knowledge or personalization capacity. Given growing concerns about AI overconfidence, hallucination, and opaque recommendation processes, transparency regarding system limitations represents a theoretically distinct and practically important design attribute. Accordingly, Study 2 extends the investigation by isolating chatbot limitation transparency and examining its direct and indirect effects on receptiveness, perceived integrity, and skepticism.
The study employed a between-subjects experimental design. Prior to the main study, a pretest (n = 66) was conducted to verify that the limitation transparency manipulation was perceived as intended and that the scenarios were clear and realistic. The results showed that the high-transparency scenario was perceived as significantly higher in chatbot limitation transparency than the low-transparency scenario (Mhigh = 4.71, SD = 0.96 vs. Mlow = 3.34, SD = 1.08, t(76) = 5.89, p < 0.001). Participants also evaluated the scenarios as sufficiently clear and realistic (Mclarity = 5.18, SD = 0.84; Mrealism = 4.88, SD = 0.91), supporting the use of the stimuli in the main experiment.
For the main study, 494 Taiwanese participants were recruited via the online platform. Among the respondents, 41.7% were male, 44.8% were aged 26–40, 87.0% were employed, and 32.8% were married. Participants in each condition were shown a vignette depicting a hypothetical interaction between them and a chatbot. They were asked to imagine opening a language model and requesting information on how to be more environmentally friendly. In the low limitation transparency condition, the chatbot provided highly confident recommendations without acknowledging any limitations. In contrast, in the high limitation transparency condition, the chatbot explicitly disclosed its epistemic constraints and uncertainty. The full experimental scenarios are provided in Appendices 3 and 4.
Thereafter, the participants responded to manipulation check items for chatbot limitation transparency (M = 3.47, SD = 1.79, α = 0.976; Hadi et al. 2024), followed by questions measuring perceived integrity (M = 4.87, SD = 0.48, α = 0.844), skepticism (M = 3.91, SD = 1.01, α = 0.926), and receptiveness to chatbot recommendations (M = 4.01, SD = 1.23, α = 0.939). All constructs were measured on a 7-point Likert scale, and the full measurement items and their sources are provided in Appendix 9. Thereafter, demographic information was collected.
To verify the effectiveness of the experimental manipulations, manipulation checks were conducted using independent samples t-tests. Results indicated that participants in the low limitation transparency condition rated limitation transparency lower than those in the high limitation transparency condition (MLowLT = 2.65, MHighLT = 5.32, t = 13.690, p < 0.001), indicating successful manipulations.
SPSS PROCESS Macro 4.2 was used to analyze the hypothesized effects. A custom model was developed to examine the effects, following the procedures set by Hayes (2022). A dummy variable for limitation transparency (0 = low limitation transparency, 1 = high limitation transparency) was created and used as the independent variable. The results (see Table 2) confirmed that high limitation transparency was associated with higher receptiveness to chatbot recommendations (β = 0.260***, CI = [0.241, 0.288]), higher perceived integrity (β = 0.828***, CI = [0.785, 0.871]), and lower skepticism (β = −0.587***, CI = [−0.431, −0.642]), supporting H2, H4, and H6. However, the results did not confirm the positive effects of perceived integrity on receptiveness to chatbot recommendations (β = 0.040, CI = [−0.119, 0.039]). Thus, H7 was not supported in study 2. The mediating role of perceived integrity in the relationship between chatbot limitation transparency and receptiveness to chatbot recommendations was not confirmed (β = 0.033, CI = [−0.035, 0.103]). H9b was not supported. On the other hand, the effect of skepticism on receptiveness to chatbot recommendations was significant (β = −0.213, CI = [−0.288, −0.137]), and the mediating effect of skepticism was significant (β = 0.125, CI = [0.171, 0.271]), supporting H8 and H10b.
The results of Study 2 showed that when chatbots acknowledge their limitations during interactions, users are more likely to accept the chatbots’ environmental recommendations. This supports the notion that transparency fosters positive user responses by signaling openness and authenticity, free from hidden biases (Chen et al. 2022; Endsley 2023; Valencia et al. 2023). However, the effect of limitation transparency on receptiveness is mediated by skepticism. While disclosing limitations enhances perceptions of integrity, these perceptions do not directly lead to greater receptiveness. This suggests that perceived integrity, which implies sincerity, honesty, and lack of bias, does not necessarily convey competence or expertise. Users may view the chatbot as truthful yet still question its ability to provide high-quality and reliable environmental advice (Hadi et al. 2024). Moreover, transparency about limitations may inadvertently emphasize the chatbot’s constrained capabilities, leading to doubts about the quality of its recommendations (Schmidt et al. 2020). Importantly, the findings confirm that limitation transparency reduces skepticism, which in turn increases receptiveness. By openly acknowledging its limitations, a chatbot may alleviate concerns of deceptive or overly confident messaging (Chen et al. 2022; Hadi et al. 2024), reinforcing its image as honest and unbiased (Binesh et al. 2024), and thereby fostering greater user receptiveness to its recommendations.
Although Study 2 demonstrates that limitation transparency reduces skepticism and increases receptiveness, it examines transparency in isolation and does not consider how transparency operates in conjunction with chatbot interactivity. In real-world AI systems, interactivity and transparency are rarely deployed independently: highly responsive chatbots may simultaneously disclose uncertainty, while fluent conversational systems may conceal epistemic limits (Binesh et al. 2024). The divergent mechanisms observed across Studies 1 and 2 (integrity-based versus skepticism-based pathways) raise an important unanswered question: how do interactivity and limitation transparency jointly shape user inferences and receptiveness? To address this conceptual gap and to test whether these attributes function as complementary or substitutive cues, Study 3 adopts a factorial design to examine their interactive effects and the underlying moderated mediation processes.
Building on Studies 1 and 2, which examined the direct effects of chatbot interactivity and limitation transparency separately, Study 3 examined their interactive effects using a 2 (interactivity: low vs. high) × 2 (limitation transparency: low vs. high) between-subjects experimental design. Prior to the main study, a pretest (n = 78) was conducted to verify that the four scenarios successfully manipulated chatbot interactivity and limitation transparency in the intended directions while remaining clear and realistic to participants. The results showed that the high-interactivity scenarios were perceived as significantly more interactive than the low-interactivity scenarios (Mhigh = 4.58, SD = 1.03 vs. Mlow = 3.31, SD = 1.16, t(154) = 5.94, p < 0.001), and that the high-transparency scenarios were perceived as significantly higher in limitation transparency than the low-transparency scenarios (Mhigh = 4.69, SD = 0.98 vs. Mlow = 3.42, SD = 1.11, t(154) = 5.71, p < 0.001). Participants also rated the scenarios as sufficiently clear and realistic (Mclarity = 5.24, SD = 0.86; Mrealism = 4.95, SD = 0.90), supporting the use of the stimuli in the main experiment.
A total of 1200 Taiwanese participants were recruited via the online platform (54.1% female, 27% aged 35–44, 79.6% employed, 44.3% married) to participate in the main study. They were rewarded with a gift card to a convenience store for their participation. Participants were randomly assigned to one of four conditions: (1) low interactivity-low limitation transparency, (2) low interactivity-high limitation transparency, (3) high interactivity-low limitation transparency, and (4) high interactivity-high limitation transparency.
In the low interactivity-low limitation transparency condition, the chatbot provided a direct response to the user’s query without inviting further interaction and without acknowledging any limitations. In the low interactivity-high limitation transparency condition, the chatbot provided a direct response without inviting further interaction, explicitly stating that it was a language model offering general guidance and was prone to limitations. In the high interactivity-low limitation transparency condition, the chatbot invited follow-up questions and additional interaction after providing its recommendation, but did not acknowledge any constraints. In the high interactivity-high limitation transparency condition, the chatbot both invited further interaction and explicitly disclosed that it was a language model offering general guidance and subject to limitations. The experimental scenarios for the four factorial conditions are presented in Appendices 5–8. The low interactivity conditions (Appendices 5 and 6) provide baseline and transparency-only manipulations, while the high interactivity conditions (Appendices 7 and 8) introduce responsive features to examine interaction effects.
After reading the scenarios, the participants responded to manipulation check items for chatbot interactivity (M = 4.00, SD = 1.28, α = 0.749) and chatbot limitation transparency (M = 4.00, SD = 1.77, α = 0.971). Thereafter, they responded to items measuring perceived integrity (M = 4.44, SD = 1.14, α = 0.925), skepticism (M = 4.30, SD = 1.52, α = 0.937) and receptiveness to chatbot recommendations (M = 4.04, SD = 0.80, α = 0.965). All constructs were measured on the 7-point Likert Scale (see Appendix 9 for detailed items and their sources).
Following prior research, this study employed the SPSS PROCESS Macro 4.2 (Model 7) to test hypothesized interaction and moderated-mediation effects (Hayes 2022; Rosnow 1985). A dummy variable capturing the joint high interactivity-high limitation transparency condition was created, with the high interactivity × high limitation transparency condition coded as 1 and all other conditions coded as 0. The dummy-coded condition was used to capture the theoretically focal combination of high interactivity and high limitation transparency and to facilitate the examination of the hypothesized moderated mediation effects.
The analysis revealed that the high interactivity-high limitation transparency condition generated significantly higher receptiveness to chatbot recommendations (β = 0.626, CI = [0.570, 0.682]), higher perceived integrity (β = 0.863, CI = [0.758, 0.968]), and lower skepticism (β = −0.256, CI = [−0.329, −0.183]). Perceived integrity was positively associated with receptiveness to chatbot recommendations (β = 0.074, CI = [0.062, 0.087]), and skepticism was negatively associated with receptiveness (β = −0.721, CI = [−0.739, −0.702]). Importantly, the indirect effects through perceived integrity (β = 0.064, CI = [0.016, 0.163]) and skepticism (β = 0.185, CI = [0.174, 0.192]) were both significant, confirming the moderated mediation effects proposed in H11a and H11b. Figure 2 and Table 3 present the PROCESS results, illustrating that the joint presence of high interactivity and high limitation transparency increased receptiveness to chatbot recommendations through enhanced perceived integrity and reduced skepticism.
The figure presents the indirect effects of the combined high-interactivity and high-limitation-transparency condition on receptiveness to chatbot recommendations, mediated by perceived integrity and skepticism.
The findings of Study 3 confirm that when chatbots simultaneously exhibit high interactivity and openly disclose their limitations, user receptiveness to chatbot recommendations substantially increases. This enhanced receptiveness can be explained by two key psychological mechanisms: increased perceived integrity and reduced skepticism.
Firstly, the interaction effect of high interactivity and high limitation transparency led to increased perceptions of perceived integrity. Users interpreted the chatbot’s openness regarding its limitations, combined with its willingness to engage interactively, as genuine, sincere, and ethically sound behavior. This aligns with prior research indicating that transparency can foster positive ethical evaluations (Binesh et al. 2024). High interactivity reinforced these perceptions by offering personalized attention, thereby demonstrating a genuine concern for user needs (Sun et al. 2024). Consequently, users felt more assured that the chatbot was honest and acting in their best interests, significantly enhancing their openness to its recommendations.
Secondly, skepticism towards chatbot recommendations decreased significantly when both interactivity and limitation transparency were high. Skepticism is often driven by suspicions of hidden motives, lack of responsiveness, or manipulative intent (Chaves and Gerosa 2021; Hadi et al. 2024). In this study, high interactivity mitigated such skepticism by creating a dynamic dialogue environment, making users feel their needs and concerns were genuinely acknowledged and addressed. Concurrently, openly communicating the chatbot’s limitations could provide explicit reassurance against exaggerated claims or hidden biases, further alleviating skepticism (Chen et al. 2022; Mozafari et al. 2021). Thus, the dual presence of these attributes synergistically fostered trust and openness towards chatbot suggestions.
Moreover, the mediating roles of perceived integrity and skepticism further illustrate how psychological perceptions shape user behavior toward chatbot interactions. Users were more receptive precisely because transparency and interactivity altered their underlying perceptions about the chatbot’s character and intentions. The results indicate the need to simultaneously implement high levels of chatbot interactivity and transparent communication about chatbot limitations.
This research explored how chatbot interactivity and limitation transparency influence users’ receptiveness to chatbot recommendations in the context of environmental communication. Three experimental studies involving Taiwanese participants provided robust evidence supporting the theory of interactive media effects (Wang and Sundar 2018). Specifically, the results highlight the critical roles of chatbot interactivity and limitation transparency. As hypothesized, chatbots with higher levels of interactivity and transparent acknowledgment of their limitations significantly enhanced user receptiveness to pro-environmental recommendations. Importantly, these effects are theoretically meaningful in environmental communication because sustainability guidance is frequently evaluated under credibility uncertainty: users cannot easily validate ecological impact claims and are primed to detect greenwashing and overstatement. Under such conditions, design cues that signal (a) responsiveness to the user and (b) epistemic humility about system limits become especially influential for receptiveness (Lewis et al. 2022; Mozafari et al. 2021; Endsley 2023; Lee and Raschke 2023).
First, the findings demonstrate that interactivity positively impacts users’ receptiveness to chatbot recommendations by enhancing perceived integrity and reducing skepticism. Interactive chatbots, by allowing dynamic, responsive, and personalized dialogue, may be perceived as fairer, more ethical and trustworthy, fostering greater acceptance of their advice (Lin 2024; Sun et al. 2024). These results extend prior research, which emphasizes that interactive communication increases message credibility and reduces perceptions of manipulative intent (Lou et al. 2022; Mozafari et al. 2021).
Second, limitation transparency proved crucial for enhancing receptiveness through reduced skepticism, providing more evidence for the positive outcomes of chatbot limitation transparency (Endsley 2023; Valencia et al. 2023). Notably, in the present non-corporate chatbot context, skepticism primarily reflected doubts about the reliability of information and overconfident AI outputs rather than suspicions of organizational self-interest, highlighting that epistemic concerns about AI accuracy represent an important source of resistance even in neutral advisory settings. When chatbots openly disclosed their limitations, users’ concerns about hidden biases or deceptive intentions were alleviated, thereby reducing skepticism and promoting greater openness toward recommendations. Interestingly, transparency directly affected skepticism rather than perceived integrity in the second study, indicating that openly acknowledging chatbot constraints primarily alleviates doubts about the chatbot’s motivations rather than enhancing its perceived expertise or competence. The differential support for H7 and H8 across studies suggests that the influence of perceived integrity and skepticism may vary depending on the specific chatbot design cue under consideration. Whereas chatbot interactivity appears to operate primarily through integrity-related evaluations, limitation transparency appears to exert its influence more strongly through skepticism reduction.
The interaction of chatbot interactivity and limitation transparency in Study 3 provided insights into how receptiveness to chatbot recommendations can be enhanced by chatbots’ interactive cues. Specifically, high interactivity and limitation transparency acted synergistically, enhancing receptiveness through increased integrity and reduced skepticism. The combination of interactive dialogue and transparent communication signals genuine intent, honesty, and responsiveness, and mitigates users’ doubts about chatbot recommendations. This finding suggests that chatbot communication outcomes are enhanced when interactivity and limitation transparency are implemented jointly rather than separately, as established by prior studies (Chaves and Gerosa 2021; Chen et al. 2022; Mozafari et al. 2021).
Nonetheless, the findings of this study also revealed complexities in the pathways linking chatbot transparency, perceived integrity, and skepticism. Neither perceived integrity nor skepticism alone consistently enhanced receptiveness, suggesting that while these factors are necessary, they may be insufficient on their own for fostering recommendation acceptance. Overall, the results indicate that chatbots must not only provide accurate information but also transparently communicate their limitations and engage users interactively. Such dual-focused designs hold greater potential for promoting meaningful and sustained pro-environmental behaviors among users.
The present work makes several theoretical contributions to the literature on AI communication and environmental messaging. First, it extends the theory of interactive media effects by emphasizing the critical roles of digital interactivity and limitation transparency in fostering positive outcomes in AI-mediated interactions. Previous studies primarily conceptualized interactivity in terms of media playability, coherence with prior messages, or content clickability, mostly within communication technologies other than chatbots (Hu and Wise 2021; Sun et al. 2024; Wang and Sundar 2018). However, interactivity also includes the capacity for conversational exchanges and explicit acknowledgment of limitations (Cheng et al. 2024; Hadi et al. 2024). These dimensions are important in the AI context, where systems can autonomously engage users in dynamic, personalized dialogues. Importantly, the present studies examine perceived message interactivity under controlled, vignette-based conditions rather than in fully dynamic, real-time conversational interaction. This approach allows the isolation of communication cues while maintaining experimental control. The findings from this study highlight the significance of both providing users with greater expressive space (i.e., interactivity) and offering transparent disclosures about system limitations. Furthermore, the results reveal that interactivity and transparency interact synergistically to shape users’ acceptance of chatbot recommendations in environmental communication.
Second, this work enriches the understanding of the psychological mechanisms underlying interactive media’s effects on user receptiveness to AI-driven communication. Specifically, it identifies the mediating roles of perceived integrity and skepticism. The findings suggest that perceived integrity primarily mediated the effects of chatbot interactivity, whereas skepticism primarily mediated the effects of limitation transparency. When both design attributes were combined, perceived integrity and skepticism jointly mediated receptiveness. These findings extend prior research on the positive outcomes of interactive media (Hadi et al. 2024; Mozafari et al. 2021) by clarifying how interactive media promotes user acceptance of chatbot recommendations in the context of environmental communication. Extant research focused on other outcomes of media interactivity such as engagement, credibility and persuasion in other contexts such as health and commerce (Hu and Wise 2021; Sun et al. 2024; Wang and Sundar 2018). The current research demonstrates that, in environmental communication contexts, reducing skepticism and enhancing perceptions of chatbot integrity can increase receptiveness to AI-generated recommendations. Environmental communication is conceptualized as a boundary condition for interactive media effects, heightening the salience of deception detection, motive attribution, and claim verifiability. Accordingly, identical design cues may function differently across domains in which claims are more readily verifiable (e.g., routine service information) or where expert judgment predominates in evaluation (e.g., medical advice), thereby outlining a clear agenda for cross-domain theoretical refinement (Longoni et al. 2019; Huang and Rust 2018).
The findings from this study offer several practical implications for organizations and developers utilizing AI chatbots in environmental communication. First, developers should prioritize creating highly interactive dialogue systems that foster two-way communication. By actively engaging users in conversations, chatbots can significantly enhance users’ perceptions of the authenticity of environmental messages, making them more receptive to pro-environmental recommendations. Companies and NGOs deploying chatbots for sustainability initiatives should ensure that their systems go beyond merely delivering generic content and instead actively invite user participation in the communication process.
Second, it is crucial for chatbots to transparently disclose their limitations to users. Managers should ensure chatbots explicitly communicate what they can and cannot do, as this can prevent users from developing unrealistic expectations. Transparent acknowledgment of chatbot limitations effectively reduces skepticism, thereby alleviating concerns about potential misinformation or manipulative intent (Chen et al. 2022; Endsley 2023). Thus, chatbot transparency serves as a vital strategy for enhancing trust and encouraging adoption of recommended environmental behaviors.
Third, our findings highlight that organizations should integrate both high interactivity and clear limitation transparency simultaneously. While each factor independently contributes positively, their combined effect is significantly stronger, suggesting that effective chatbot design should strategically incorporate both attributes. Organizations should thus carefully balance these elements and ensure that their chatbots remain interactive while clearly articulating their functional boundaries.
Despite the insights provided by this investigation, several limitations remain that could be addressed in future research. First, the research context was confined to environmental communication with Taiwanese samples, so the generalizability of the findings to other cultural contexts or communication domains is uncertain. Future studies should explore whether these results extend to different cultures and communication areas, such as health or financial recommendations, where trust dynamics may differ significantly (Huang and Rust 2018; Longoni et al. 2019).
Second, although the stimulus materials were designed to reflect realistic and commonly encountered environmental advice, the use of a single standardized scenario per condition may limit the generalizability of the findings. Future research should examine multiple scenarios, recommendation types, and contextual variations to assess the robustness of the observed effects across different environmental domains and interaction settings.
Third, the current research focused primarily on short-term effects and user receptiveness to immediate chatbot recommendations. Longitudinal research would be beneficial in understanding whether initial receptiveness translates into sustained behavioral change over time. Such studies could examine the long-term impact of interactive and transparent chatbot interactions on the adoption of pro-environmental behaviors, providing deeper insights into how chatbots’ interactive features enhance behavioral outcomes over time.
Fourth, the experimental studies relied on hypothetical scenarios to assess user reactions. Real-world settings may introduce additional complexities that influence user acceptance of chatbot recommendations. Future research could employ field experiments or real chatbot interactions to validate these findings, thereby offering greater ecological validity and deeper insights into the real-world effectiveness of chatbots in environmental communication (Cheng et al. 2024; Valencia et al. 2023).
Finally, although the present research conceptualizes interactivity as a positive design attribute that enhances receptiveness and trust, persuasive interactivity may also be deployed in manipulative or ethically problematic ways. Highly responsive and adaptive conversational systems can potentially amplify persuasion pressure, obscure persuasive intent, or exploit users’ cognitive and emotional vulnerabilities, particularly in domains involving moral or normative appeals such as environmental communication (Chen et al. 2022). The current studies did not examine these potential dark-side effects of interactive persuasion. Future research should therefore investigate when and how interactivity may undermine user autonomy, elicit undue compliance, or contribute to ethical risks, as well as identify design principles that balance persuasive effectiveness with transparency, user agency, and ethical safeguards.
This research examines how chatbot interactivity and limitation transparency shape users’ receptiveness to chatbot recommendations in environmental communication. Across three experimental studies, the findings show that chatbot interactivity increases receptiveness primarily through enhanced perceived integrity, whereas transparency about limitations reduces skepticism and thereby increases receptiveness. Study 3 further demonstrates that combining high interactivity with high limitation transparency produces the strongest effects by simultaneously strengthening integrity and lowering skepticism.
These findings extend the theory of interactive media effects by showing that responsiveness and transparent disclosure function as complementary design cues in AI-mediated environmental communication. They also offer practical guidance by indicating that chatbots are likely to be more effective when they are both engaging and explicit about their limitations. As conversational AI becomes more widely used in sustainability communication, aligning interactivity with transparency will be essential to fostering credible, responsible message acceptance. Together, these findings suggest that the future effectiveness of AI-mediated environmental communication depends not only on creating engaging interactions but also on fostering trust by transparently acknowledging system limitations.
All data generated or analyzed in this study have been included in the manuscript and the Supplementary Files.
Ahadzadeh AS, Ong FS, Wu SL (2021) Social media skepticism and belief in conspiracy theories about COVID-19: the moderating role of the dark triad. Curr Psychol. https://doi.org/10.1007/s12144-021-02198-1
Aljukhadar M, Senecal S, Ouellette D (2010) Can the media richness of a privacy disclosure enhance outcome? A multifaceted view of trust in rich media environments. Int J Electron Commer 14(4):103–126. https://doi.org/10.2753/JEC1086-4415140404
Article Google Scholar
Binesh N, Rippee R, Belarmino A (2024) User segmentation in human–robot interactions: insights from sports betting patrons using diffusion of innovation theory. Curr Issues Tour 0(0):1–21. https://doi.org/10.1080/13683500.2024.2372001
Article Google Scholar
Chaves AP, Gerosa MA (2021) How should my chatbot interact? A survey on human-chatbot interaction design. Int J Hum Comput Interact 37(8):729–758. https://doi.org/10.1080/10447318.2020.1841438
Article Google Scholar
Chen Q, Gong Y, Lu Y, Tang J (2022) Classifying and measuring the service quality of AI chatbot in frontline service. J Bus Res 145:552–568. https://doi.org/10.1016/j.jbusres.2022.02.088
Article Google Scholar
Cheng Y, Wang Y, Lee J (2024) Using a chatbot to combat misinformation: exploring gratifications, chatbot satisfaction and engagement, and relationship quality. Int J Hum Comput Interact 0(0):1–13. https://doi.org/10.1080/10447318.2024.2344149
Article Google Scholar
Endsley MR (2023) Supporting human-AI teams:transparency, explainability, and situation awareness. Comput Hum Behav 140: 107574. https://doi.org/10.1016/j.chb.2022.107574
Article Google Scholar
Følstad A, Nordheim CB, Bjørkli CA (2018) What makes users trust a chatbot for customer service? An exploratory interview study. In: Bodrunova SS (ed) Internet science. Springer International Publishing, 194–208. https://doi.org/10.1007/978-3-030-01437-7_16
Goh SK, Balaji MS (2016) Linking green skepticism to green purchase behavior. J Clean Prod 131:629–638. https://doi.org/10.1016/j.jclepro.2016.04.122
Article Google Scholar
Hadi ZA, Siregar DA, Wijaya GST, Handayani PW, Harahap NC (2024) The influence of transparency, anthropomorphism, and positive politeness on chatbots for service recovery in E-health applications. Cogent Soc Sci 10(1):2415534. https://doi.org/10.1080/23311886.2024.2415534
Article Google Scholar
Hayes AF (2022) Introduction to mediation, moderation, and conditional process analysis: third edition: a regression-based approach, 3rd edn. Guilford Press. https://www.guilford.com/books/Introduction-to-Mediation-Moderation-and-Conditional-Process-Analysis/Andrew-Hayes/9781462549030
Hu X, Wise K (2021) How playable ads influence consumer attitude: exploring the mediation effects of perceived control and freedom threat. J Res Interact Mark 15(2):295–315. https://doi.org/10.1108/JRIM-12-2020-0269
Article Google Scholar
Huang M-H, Rust RT (2018) Artificial intelligence in service. J Serv Res 21(2):155–172. https://doi.org/10.1177/1094670517752459
Article Google Scholar
Jiang Y, Yang X, Zheng T (2023) Make chatbots more adaptive: dual pathways linking human-like cues and tailored response to trust in interactions with chatbots. Comput Hum Behav 138: 107485. https://doi.org/10.1016/j.chb.2022.107485
Article Google Scholar
Johnson M (2023) Ecosia debuts AI chatbot for eco-friendly searches. KillerStartups. https://www.killerstartups.com/ecosia-debuts-ai-chatbot-for-eco-friendly-searches/
Lee MT, Raschke RL (2023) Stakeholder legitimacy in firm greening and financial performance: what about greenwashing temptations? J Bus Res 155: 113393. https://doi.org/10.1016/j.jbusres.2022.113393
Article Google Scholar
Lee SS, Johnson BK (2022) Are they being authentic? The effects of self-disclosure and message sidedness on sponsored post effectiveness. Int J Advert 41(1):30–53. https://doi.org/10.1080/02650487.2021.1986257
Article Google Scholar
Lee SY, Kim Y, Kim Y (2021) Engaging consumers with corporate social responsibility campaigns: the roles of interactivity, psychological empowerment, and identification. J Bus Res 134:507–517. https://doi.org/10.1016/j.jbusres.2021.05.062
Article Google Scholar
Lewis N, Rossmann C, de Bruijn G-J, Martinez LS (2022) Dual process models and information engagement: testing effects of seeking, scanning, and trust in sources on attitudes toward marijuana. J Commun 72(1):59–82. https://doi.org/10.1093/joc/jqab024
Article Google Scholar
Lin J-S (2024) Examining the interplay of chatbot interactivity and anthropomorphic design for enhancing brand relationships in service businesses. Int J Advert 0(0):1–32. https://doi.org/10.1080/02650487.2024.2410559
Article Google Scholar
Longoni C, Bonezzi A, Morewedge CK (2019) Resistance to medical artificial intelligence. J Consum Res 46(4):629–650. https://doi.org/10.1093/jcr/ucz013
Article Google Scholar
Lou C, Kang H, Tse CH (2022) Bots vs. humans: how schema congruity, contingency-based interactivity, and sympathy influence consumer perceptions and patronage intentions. Int J Advert 41(4):655–684. https://doi.org/10.1080/02650487.2021.1951510
Article Google Scholar
Mozafari N, Weiger WH, Hammerschmidt M (2021) Trust me, I’m a bot—repercussions of chatbot disclosure in different service frontline settings. J Serv Manag 33(2):221–245. https://doi.org/10.1108/JOSM-10-2020-0380
Article Google Scholar
Pizzi G, Scarpi D, Pantano E (2021) Artificial intelligence and the new forms of interaction: who has the control when interacting with a chatbot? J Bus Res 129:878–890. https://doi.org/10.1016/j.jbusres.2020.11.006
Article Google Scholar
Rosnow RL (1985) Contrast analysis: focused comparisons in the analysis of variance. Cambridge University Press
Santos C, Coelho A, Marques A (2024) A systematic literature review on greenwashing and its relationship to stakeholders: state of art and future research agenda. Manag Rev Q 74(3):1397–1421. https://doi.org/10.1007/s11301-023-00337-5
Article Google Scholar
Schmidt P, Biessmann F, Teubner T (2020) Transparency and trust in artificial intelligence systems. J Decis Syst 29(4):260–278. https://doi.org/10.1080/12460125.2020.1819094
Article Google Scholar
Su Y, Lee DKL, Xiao X, Li W, Shu W (2021) Who endorses conspiracy theories? A moderated mediation model of Chinese and international social media use, media skepticism, need for cognition, and COVID-19 conspiracy theory endorsement in China. Comput Hum Behav 120: 106760. https://doi.org/10.1016/j.chb.2021.106760
Article Google Scholar
Sun Y, Chen J, Sundar SS (2024) Chatbot ads with a human touch: a test of anthropomorphism, interactivity, and narrativity. J Bus Res 172: 114403. https://doi.org/10.1016/j.jbusres.2023.114403
Article Google Scholar
Sun Y, Zhong Y, Zhang Z, Wang Y, Zhu M (2024) How technical features of virtual live shopping platforms affect purchase intention: based on the theory of interactive media effects. Decis Support Syst 180: 114189. https://doi.org/10.1016/j.dss.2024.114189
Article Google Scholar
Valencia OAG, Suppadungsuk S, Thongprayoon C, Miao J, Tangpanithandee S, Craici IM, Cheungpasitporn W (2023) Ethical implications of chatbot utilization in nephrology. J Pers Med 13(9):9. https://doi.org/10.3390/jpm13091363
Wang R, Sundar SS (2018) How does parallax scrolling influence user experience? A test of TIME (theory of interactive media effects). Int J Hum Comput Interact 34(6):533–543. https://doi.org/10.1080/10447318.2017.1373457
Article Google Scholar
Wanner J, Herm L-V, Heinrich K, Janiesch C (2022) The effect of transparency and trust on intelligent system acceptance: evidence from a user-based study. Electron Mark 32(4):2079–2102. https://doi.org/10.1007/s12525-022-00593-5
Article Google Scholar
World Wildlife Fund (2024) TNRC Blog | “Justina del Mar,” a virtual ally to prevent corruption in the artisanal fishing sector. World Wildlife Fund. https://www.worldwildlife.org/pages/tnrc-blog-justina-del-mar-a-virtual-ally-to-prevent-corruption-in-the-artisanal-fishing-sector
Xie Y, Peng S (2009) How to repair customer trust after negative publicity: the roles of competence, integrity, benevolence, and forgiveness. Psychol Mark 26(7):572–589. https://doi.org/10.1002/mar.20289
Article Google Scholar
Yin C-Y, Bi N, Chen Y (2022) Cut-throat competition or justice? Consumers’ responses to competitor whistleblowing from the perspective of consumer skepticism. J Mark Manag 38(7–8):633–660. https://doi.org/10.1080/0267257X.2021.1963310
Article Google Scholar
Download references
Bachelor Program in Digital Media and Technology, Tzu Chi University, Hualien City, Taiwan
Chi-Horng Liao
Department of Communication Studies, Tzu Chi University, Hualien City, Taiwan
Chi-Horng Liao
Department of Information Management, National Changhua University of Education, Changhua, Taiwan
Hope Wilfred Banda
PubMed Google Scholar
PubMed Google Scholar
CHL conceptualized the study, designed and conducted the research, performed data analysis, developed the theoretical framework, and wrote and approved the final manuscript. HWP contributed to the original draft writing, methodology development, conceptualization, data analysis, and manuscript review and editing. All authors reviewed and revised the manuscript.
Correspondence to Chi-Horng Liao.
The authors declare no competing interests.
The study involved anonymous, non-interactive, and non-invasive data collection, and no personally identifiable information was obtained. The research did not target vulnerable populations (e.g., minors, individuals in institutions, or individuals unable to provide informed consent), and the procedures posed no more than minimal risk to participants. Accordingly, this study met the criteria for exemption from institutional ethics review under Taiwan’s Human Subjects Research Act, as specified in the official announcement “Categories of Human Subject Research Exempt from Ethics Review” (Ministry of Health and Welfare Announcement No. 1010265075, issued on July 5, 2012).
Informed consent was obtained from all participants prior to the administration of the survey. Participants were informed about the purpose of the study, the voluntary nature of participation, and their right to withdraw at any time without penalty. All responses were collected anonymously, and no personally identifiable information was recorded. By proceeding with the survey, participants indicated their informed consent to participate.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Reprints and permissions
Liao, CH., Wilfred Banda, H. Talking green: how chatbot interactivity and transparency foster user receptivity to chatbot recommendations in environmental communication. Humanit Soc Sci Commun 13, 1716 (2026). https://doi.org/10.1057/s41599-026-07965-8
Download citation
Received:
Accepted:
Published:
Version of record:
DOI: https://doi.org/10.1057/s41599-026-07965-8
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.
Provided by the Springer Nature SharedIt content-sharing initiative
Advertisement
Humanities and Social Sciences Communications (Humanit Soc Sci Commun)
ISSN 2662-9992 (online)
© 2026 Springer Nature Limited